Survival Benefit of Corticosteroid Replacement in Critically Ill Patients: One Treatment Effect or Two?

Survival benefit of corticosteroid replacement in critically ill patients: one treatment effect or two? Background: In the comparison of placebo versus corticosteroid replacement therapy (CRT) among ill patients, the treatment effect (difference in 28-day mortality rate between placebo and CRT) tends to vary according to mortality in the placebo group. In this context, we treat 28-day mortality in the placebo group as a surrogate for severity of illness [1]. Hypothesis: In the present era of physiologic CRT, critical illness related corticosteroid insufficiency (CIRCI) may be defined operationally by the difference in clinical outcomes, such as 28-day mortality, between placebo and CRT treatment groups. In this simulation analysis, we examined the hypothesis that treatment effects of CRT observed in various randomized, placebo-controlled clinical trials may be explained by heterogeneous population consisting of two strata, S1 and S2, having discrete conditional probabilities of mortality when treated with CRT vs. placebo. Methods: Using published Randomized Controlled Trial (RCT) data, the relationship between treatment effect and severity of illness was analyzed using a least squares solution weighted by sample sizes. A probability model included likelihoods for outcomes (28-day mortality with placebo vs. CRT) in two population strata: (i) a minority stratum (S1) having treatment effect > 0 and (ii) a majority stratum (S2) having a treatment effect ≤ 0. In a simple simulation scenario we varied the treatment effect in S1 as function of the placebo group mortality in S2 in order to fit the relationship obtained from published RCT data. The potential impact of sensitivity and specificity on diagnostic tests to identify S1 was also addressed. Results: A hyperbolic function (y = [Mx/(K+x)] - A) provided a good fit for published RCT trials of physiologic CRT (treatment effect [y] vs. placebo group mortality rate [x]). Pooling results for S1 and S2 having population frequencies of 0.2 and 0.8, respectively, approximated the observed hyperbolic function of treatment effects and severity of illness. A significant treatment effect was maintained when S1 was identified by diagnostic tests having 75% sensitivity-specificity. Conclusions: Population heterogeneity has not been excluded as an explanation for the relatively modest treatment effect of CRT in critically ill patients. Given the possibility of harm of CRT in the majority population (S2), our model further suggests that clinical or laboratory measures that would distinguish S1 and S2 with reasonable accuracy would significantly improve clinical outcomes and reduce numbers needed to treat. Our findings also support the more general hypothesis that the treatment effects of CRT vary according to severity of illness (1). References: 1.Briegel, J., V. Huge, and P. Mohnle, Hydrocortisone in septic shock: all the questions answered? J. Thorac. Dis, 2018. 10(Suppl 17): p. S1962-S1965.